Riboflavin, vitamin B.sub.2, is the precursor of flavin mononucleotide (FMN) and flavin adenine dinucleotide (FAD), essential cofactors for a number of mainstream metabolic enzymes that mediate hydride, oxygen, and electron transfer reactions. Riboflavin-dependent enzymes include succinate dehydrogenase, NADH dehydrogenase, ferredoxin-NADP.sup.+ oxidoreductase, acyl-CoA dehydrogenase, and the pyruvate dehydrogenase complex. Consequently, fatty acid oxidation, the TCA cycle, mitochondrial electron-transport, photosynthesis, and numerous other cellular processes are critically dependent on either FMN or FAD as prosthetic groups. Other notable flavoproteins include glutathione reductase, glycolate oxidase, P450 oxido-reductase, squalene epoxidase, dihydroorotate dehydrogenase, and .alpha.-glycerophosphate dehydrogenase. Genetic disruption of riboflavin biosynthesis in E. coli (Richter et al., J. Bacteriol.174:4050-4056 (1992)) and S. cerevisiae (Santos et al., J. Biol. Chem. 270:437-444 (1995)) results in a lethal phenotype that is only overcome by riboflavin supplementation. This is not surprising, considering the ensemble of deleterious pleiotropic effects that would occur with riboflavin deprivation.
Riboflavin is synthesized by plants and numerous microorganisms, including bacteria and fungi (Bacher, A., Chemistry and Biochemistry of Flavoproteins (Muller, F., ed.) vol. 1, pp. 215-259, Chemical Rubber Co., Boca Raton, FL (1990)). Since birds, mammals, and other higher organisms are unable to synthesize the vitamin and, instead, rely on its dietary ingestion to meet their metabolic needs, the enzymes that are responsible for riboflavin biosynthesis are potential targets for future antibiotics, fungicides, and herbicides. Moreover, it is possible that the distantly-related plant and microbial enzymes have distinct characteristics that could be exploited in the development of potent organism-specific inhibitors. Thus, a detailed understanding of the structure, mechanism, kinetics, and substrate-binding properties of the riboflavin biosynthetic enzyme(s), from plants for example, would serve as a starting point for the rational design of chemical compounds that might be useful as herbicides. Having the authentic plant protein(s) in hand would also provide a valuable tool for the in vitro screening of chemical libraries in search of riboflavin biosynthesis inhibitors.
Bacterial and fungal riboflavin biosynthesis has been intensively studied for more than four decades (For recent reviews, see Bacher, A., Chemistry and Biochemistry of Flavoproteins (Moller, F., ed.) vol. 1, pp. 215-259 and 293-316 Chemical Rubber Co., Boca Raton, FL (1990)). The synthetic pathway consists of seven distinct enzyme catalyzed reactions, with guanosine 5'-triphosphate (GTP) and ribulose 5-phosphate the ultimate precursors. While the second and third steps of riboflavin biosynthesis occur in opposite order in bacteria and fungi, the remaining pathway intermediates are identical in both microorganisms. Structurally and mechanistically, the last two reactions in the pathway, namely, those catalyzed by 6,7-dimethyl-8-ribityllumazine synthase (LS) and riboflavin synthase (RS), are best characterized. In B. subtilis, these two enzymes are physically associated with each other in a huge spherical particle with a combined molecular mass of about 1 MDa (Bacher et al., J. Biol. Chem. 255:632-637 (1980); Ritsert et al., J. Mol. Biol. 253, 151-167 (1995); Bacher et al., Biochem. Soc. Trans. 24(1):89-94 (1996)); the X-ray structure of the bifunctional protein complex has been determined at 3.3 angstrom resolution (Ladenstein et al., J. Mol. Biol 203:1045-1070). The LS/RS complex consists of 60 LS subunits that are organized into 12 pentamers to form a hollow icosahedral capsid. Encaged in the central core of this structure resides a single molecule of RS, a trimer of three identical subunits. Kinetic studies reveal that the compartmentation of the two enzymes within the complex improves the overall catalytic efficiency of riboflavin production at low substrate concentrations, presumably via "substrate channeling" (Kis et al., J. Biol. Chem. 270:16788-16795 (1995)). Although a bifunctional LS/RS complex has not been observed in other microorganisms, it was recently shown that the native E. coli LS also exists in vivo as a hollow icosahedral capsid of 60 identical subunits (Mortl et al., J. Biol. Chem. 271:33201-33207 (1996)).
LS, the penultimate enzyme of riboflavin biosynthesis, catalyzes the condensation of 3,4-dihydroxy-2-butanone 4-phosphate with 4-ribitylamino-5-amino-2,6-dihydroxypyrimidine (RAADP) to yield 1 mol each of orthophosphate and 6,7-dimethyl-8-(1'-D-ribityl)-lumazine (DMRL). The latter is the immediate precursor of riboflavin. LS-encoding genes have been cloned from numerous microorganisms, including E. coli (Taura et al., Mol. Gen. Genet. 234:429-432 (1992)), A. pleuropneumoniae (Fuller et al., J. Bacteriol. 177:7265-7270 (1995)), P. phosphoreum (Lee et al., J. Bacteriol. 176:2100-2104 (1994)), B. subtilis (Mironov et al., Dokl. Akad. Nauk SSSR 305:482-487 (1989)), and S. cerevisiae (Garcia-Ramirez et al., J. Biol. Chem. 270:23801-23807 (1995)). In all cases, the subunit molecular mass of the LS gene product is small, ranging in size from .about.16-17 kDa. While the various LS homologs all share certain structural features in common, their overall homology at the primary amino acid sequence level is rather poor. For example, as determined with the Genetics Computer Group Gap program (Wisconsin Package Version 9.0, Genetics Computer Group (GCG), Madison, Wisc.), the E. coli LS is only 58%, 65%, 53%, and 36% identical to the homologous proteins of A. pleuropneumoniae, P. phosphoreum, B. subtilis and S. cerevisiae, respectively. Indeed, pairwise comparisons of these five proteins reveal that the two most similar homologs share only 72% identity.
The terminal step of riboflavin biosynthesis is mediated by RS. This enzyme catalyzes the dismutation of two molecules of DMRL to yield 1 mol of riboflavin and RAADP. That the latter product is also one of the substrates of LS explains in part the enhanced catalytic efficiency of the B. subtilis LS/RS complex noted above. Although the crystal structure of RS remains to be determined, it is surmised that the native bacterial (Bacher et al., J. Biol. Chem. 255:632-637 (1980)) and fungal (Santos et al., J. Biol. Chem. 270:437-444 (1995)) proteins are trimers, each consisting of three identical .about.25 kDa subunits. To date, RS has only been cloned from about a dozen microorganisms, and all of the species that have been examined exhibit marked internal homology in their N-terminal and C-terminal domains (Schott et al., J. Biol. Chem. 265:4204-4209 (1990); Santos et al., J. Biol. Chem. 270:437-444 (1995)). Based on these observations, it has been suggested that the two halves of the RS protomer have arisen through gene duplication, and that each contains a substrate-binding site for DMRL. Despite this structural similarity, however, the overall sequence homology of the various RS proteins is extremely limited. Thus, the E. coli RS protein is only 32%, 36%, 35%, and 31% identical to its counterparts in S. cerevisiae, P. phosphoreum., B. subtilis, and P. leiognathi; the GenBank accession numbers for the latter four proteins are Z21621, L 11391, X51510 and M90094, respectively.
With the exception of GTP cyclohydrolase II, the first committed enzyme of riboflavin biosynthesis, virtually nothing is known about the riboflavin biosynthetic machinery of higher plants. The gene for this protein was recently cloned from an arabidopsis cDNA library (Kobayashi et al., Gene 160:303-304 (1995)). The protein sequence of the cloned plant gene is only 37-58% identical to the homologous proteins from E. coli, B. subtilis, P. leiognathi, and P. phosporeum. While full-length cDNA sequences have not been reported for any other plant riboflavin biosynthetic enzyme, the GenBank database contains two ESTs (Expressed Sequence Tags) that potentially correspond to plant LS genes. One of these is from castor bean and the other is from arabidopsis.
The castor bean cDNA clone (GenBank accession number T15152; van de Loo et al., Plant Physiol. 108:1141-1150 (1995)) is truncated at its 5' end, and is missing DNA corresponding to at least 60 N-terminal amino acid residues. The arabidopsis cDNA clone (GenBank accession number Z34233; direct submission) was identified through a BLAST (Basic Local Alignment Search Tool; Altschul et al., J. Mol. Biol. 215:403-410 (1990)) search using the TBLASTN algorithm provided by the National Center for Biotechnology Information (NCBI). The query sequence for the BLAST search was the translated E. coli LS gene (GenBank accession number X64395) and the probability score for similarity to the arabidopsis EST was P=0.45. Unfortunately, the portion of the cDNA insert that was sequenced contained only the last 26 C-terminal amino acid residues of the protein, so it is not known whether it is a partial or full-length cDNA clone. Since neither of these clones possess a polyA tail, it is possible that they reflect contaminating microbial DNA that was introduced at some point during the preparation of the cDNA libraries.
In contrast to LS, BLAST searches failed to identify any plant DNA sequences in the GenBank database with significant primary amino acid sequence homology to either E. coli or yeast RS. However, RS activity has been detected in extracts from various plant species (Plaut, G., Metabolic Pathways (Greenberg, D.M., ed.), vol II, p. 673, Academic Press, New York, (1961)), and partial purification of the spinach homolog has been described (Mitsuda et al., Methods Enzymol. 18b:539-543 (1970)).
From the foregoing discussion, it is apparent that too little is known about plant LS or RS genes/proteins and their relationship to known microbial homologs to allow isolation of LS- or RS-encoding genes from any plant species using most classical approaches. The latter include hybridization probing of cDNA libraries with homologous or heterologous genes, PCR-amplification of the gene of interest using oligionucleotide primers corresponding to conserved amino acid sequence motifs, and/or immunological detection of expressed cDNA inserts in microbial hosts. Unfortunately, these techniques would not be expected to be very useful for the isolation of plant LS or RS genes, since they all heavily rely on the presence of significant structural similarity (i.e., DNA or amino acid sequence) with known proteins and genes that have the same function. Given the observation that LS and RS proteins are both so poorly conserved, even amongst microorganisms, it is highly unlikely that the known microbial homologs would share significant structural similarities with their counterparts in higher plants.
An alternative approach that has been used to clone biosynthetic genes in other metabolic pathways from higher eucaryotes is through complementation of microbial mutants that are deficient in the enzyme activity of interest. Since this strategy relies only on the fiinctional similarity between the protein encoded for by the disrupted host gene and the target gene of interest, it is ideally suited for cloning structurally dissimilar proteins that catalyze the same reaction. For functional complementation, a cDNA library is constructed in a vector that can direct the expression of the cDNA in the microbial host. The plasmid library is then introduced into the mutant microbe, and colonies are selected that are no longer phenotypically mutant. Indeed, the LS (Garcia-Ramirez et al., J. Biol. Chem. 270:23801-23807 (1995)) and RS (Santos et al., J. Biol. Chem. 270:437-444 (1995) of yeast, and arabidopsis GTP cyclohydrolase II (Kobayashi et al, Gene 160:303-304 (1995)) were all cloned through finctional complementation of microbial riboflavin auxotrophs. This strategy has also worked for isolating genes from higher eucaryotes that are involved in other metabolic pathways, including lysine biosynthesis (Frisch et al., Mol. Gen. Genet. 228:287-293 (1991)), purine biosynthesis (Aimi et al., J. Biol. Chem. 265:9011-9014 (1990)), and tryptophan biosynthesis (Niyogi et al., Plant Cell 5:1011-1027 (1993)), and has also been successfully employed in the isolation of various plant genes including glutamine synthetase (Snustad et al., Genetics 120:1111-1124 (1988)), pyrroline-5-carboxylate reductase (Delauney et al., Mol. Genet. 221:299-305 (1990)), dihydrodipicolinate synthase (Frisch et al., Mol. Gen. Genet. 228:287-293 (1991)), 3-isopropylmalate dehydrogenase (Ellerstrom et al., Plant Mol. Biol. 18:557-566 (1992)), and dihydroorotate dehydrogenase (Minet et al., Plant J. 2:417-422 (1992)).
Despite the obvious attractive features of cloning by functional complementation, there are several reasons why this approach might not work when applied to the higher plant LS and RS genes. First, the eucaryotic cDNA sequence might not be expressed at adequate levels in the mutant microbe for a variety of reasons, including differences in preferred codon usage. Second, the cloned eucaryotic gene might not produce a functional polypeptide, if for instance, enzyme activity requires a post-translational modification, such as acetylation, glycosylation, or phosphorylation that is not carried out by the microbial host. Third, the heterologous plant protein might be lethal to the host, thus rendering its expression impossible. Fourth, the eucaryotic protein might fail to achieve its native conformation in the foreign microbial environment, due to folding problems, inclusion body formation, or various other reasons. It is also possible that the higher plant LS and RS enzymes are nuclear-encoded proteins that are posttranslationally targeted to chloroplasts, mitochondrial, or some other organelle that is not present in the microbial host. If this were the case and proteolytic removal of the organellar targeting sequence was required for enzyme activity, cloning these genes by functional complementation would not be possible.